https://lens.taqlio.com/ https://lens.taqlio.com/bibliography https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/current-capabilities https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/foundation-models https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/defining-and-measuring-agi https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/leveraging-scale https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/forecasting-timelines https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/takeoff https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/appendix-forecasting https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/appendix-takeoff https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/appendix-expert-surveys https://lens.taqlio.com/chapter/ai-safety-atlas-capabilities/appendix-discussion-on-llms https://lens.taqlio.com/chapter/ai-safety-atlas-risks https://lens.taqlio.com/chapter/ai-safety-atlas-risks/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-risks/risk-decomposition https://lens.taqlio.com/chapter/ai-safety-atlas-risks/dangerous-capabilities https://lens.taqlio.com/chapter/ai-safety-atlas-risks/misuse-risks https://lens.taqlio.com/chapter/ai-safety-atlas-risks/misalignment-risks https://lens.taqlio.com/chapter/ai-safety-atlas-risks/systemic-risks https://lens.taqlio.com/chapter/ai-safety-atlas-risks/risk-amplifiers https://lens.taqlio.com/chapter/ai-safety-atlas-risks/conclusion https://lens.taqlio.com/chapter/ai-safety-atlas-risks/appendix-quantifying-existential-risks https://lens.taqlio.com/chapter/ai-safety-atlas-risks/appendix-forecasting-scenarios https://lens.taqlio.com/chapter/ai-safety-atlas-strategies https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/definitions https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/misuse-prevention-strategies https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/agi-safety-strategies https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/asi-safety-strategies https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/socio-technical-strategies https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/combining-strategies https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/challenges https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/conclusion https://lens.taqlio.com/chapter/ai-safety-atlas-strategies/appendix-long-term-questions https://lens.taqlio.com/chapter/ai-safety-atlas-governance https://lens.taqlio.com/chapter/ai-safety-atlas-governance/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-governance/governance-problems https://lens.taqlio.com/chapter/ai-safety-atlas-governance/compute-governance https://lens.taqlio.com/chapter/ai-safety-atlas-governance/systemic-challenges https://lens.taqlio.com/chapter/ai-safety-atlas-governance/governance-architectures https://lens.taqlio.com/chapter/ai-safety-atlas-governance/implementation https://lens.taqlio.com/chapter/ai-safety-atlas-governance/conclusion https://lens.taqlio.com/chapter/ai-safety-atlas-governance/appendix-data-governance https://lens.taqlio.com/chapter/ai-safety-atlas-governance/appendix-national-governance https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/benchmarks https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/evaluated-properties https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/evaluation-techniques https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/evaluation-frameworks https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/dangerous-capability-evaluations https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/dangerous-propensity-evaluations https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/control-evaluations https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/evaluation-design https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/limitations https://lens.taqlio.com/chapter/ai-safety-atlas-evaluations/conclusion https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming/reinforcement-learning https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming/optimization https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming/specification-gaming https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming/learning-from-imitation https://lens.taqlio.com/chapter/ai-safety-atlas-specification-gaming/learning-from-feedback https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/multi-objective-generalization https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/learning-dynamics https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/goal-directedness https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/scheming https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/detection https://lens.taqlio.com/chapter/ai-safety-atlas-goal-misgeneralization/mitigations https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/introduction https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/oversight https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/task-decomposition https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/process-oversight https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/iterated-amplification https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/debate https://lens.taqlio.com/chapter/ai-safety-atlas-scalable-oversight/weak-to-strong-w2s